First, the word. Exposure means AI can already do a meaningful chunk of what a job involves. It does not mean "you're fired tomorrow." Hold that distinction — most of the panic online comes from blurring it.
Higher exposure (often)
- Desk and screen work — reports, email, scheduling, data entry, standard customer replies.
- Routine cognitive tasks — templates, summaries, first drafts, basic analysis.
- Junior white-collar rungs — the work that used to train graduates before they earned senior judgement.
- Generic output — boilerplate copy, basic graphics, first-line support scripts.
- Parts of finance, law, marketing and software — anywhere the output is essentially language-shaped.
Lower exposure (often — but not zero)
- Skilled trades on site — plumber, tattooer, electrician. Robots lag badly here. But the booking and admin around the trade isn't safe at all.
- Physical care built on human trust — the nurse's touch, childcare. The admin around it still automates.
- Licensed, on-site judgement roles — protected until the robots-plus-agents combo properly matures.
Find yourself in this
- Tradesperson — the hands-on work is safe longer; your quotes, bookings, DMs and marketing are not.
- Office admin — email, scheduling, standard replies. Squarely in the mix.
- Shop owner — you're usually all of the above, plus spreadsheets and customer messages.
- Freelancer — drafts, proposals, research. Language-shaped from dawn till dusk.
Here's the uncomfortable maths. Loads of jobs are roughly 80% routine, 20% judgement. An employer can keep one person plus AI where they used to need three. Which is why the national unemployment figure can read "fine" on the news while your specific role quietly vanishes.
If you earn your wage sitting at a computer, take this one seriously.
Continue — the numbers, with citations.